Skip to main content

Build with Cogneris

New · Cogneris MCP server

Integrate Cogneris by asking your AI assistant.

Connect VS Code, Visual Studio, Cursor or Claude Code to the Cogneris MCP server and describe what you want — "extract the data from our supplier invoices", "create a sandbox API key", "send me a webhook when a document is processed". The assistant reads the real API contract, creates the key and the webhook on your tenant, and writes the code in your project.

Connect in two minutesSee the prompts
Cogneris MCP: Cogneris in VS Code: MCP, API keys, Python code and webhooks Watch on YouTube

Connect in two minutes​

1. Generate a developer token. In the Cogneris platform, open Vision → MCP. Choose a name, a validity of 30 or 90 days and the scopes: mcp.read and mcp.write are ticked by default; add apikeys.manage and webhooks.manage if the assistant should create API keys and webhooks for you. Copy the token — it is shown once.

2. Add the server to your editor. Use the URL of your region: https://mcp-eu.cogneris.ai/ (Europe, the default) or https://mcp-us.cogneris.ai/ (United States). The Vision → MCP page shows the right one for your tenant.

Create .vscode/mcp.json in your project. VS Code asks for the token once and keeps it in its secret storage, so this file is safe to commit.

.vscode/mcp.json
{
  "inputs": [
    {
      "id": "cogneris-token",
      "type": "promptString",
      "password": true,
      "description": "Cogneris developer token"
    }
  ],
  "servers": {
    "cogneris": {
      "type": "http",
      "url": "https://mcp-eu.cogneris.ai/",
      "headers": { "Authorization": "Bearer ${input:cogneris-token}" }
    }
  }
}

Open Copilot Chat, switch it to Agent mode, and the Cogneris tools appear in the tool picker.

3. Ask. Start with "Who am I connected as on Cogneris?" — the assistant calls cogneris_whoami and answers with your tenant, user, scopes and the API base URL your code should use. Then ask for what you are building.

Full setup, browser sign-in and troubleshooting: Connect your editor to the MCP server.

What you can ask for​

Integrate extraction

Add a Cogneris extraction client to this project. When a PDF lands in the uploads folder, send it to the Cogneris extraction API and save the extracted fields as JSON. Read the key from COGNERIS_API_KEY.

What the assistant does →

Create an API key

Create a sandbox Cogneris API key called "invoice-importer-dev" with only the scopes this code needs, and put it in my .env.

What the assistant does →

Create a webhook

Register a Cogneris webhook at https://api.example.com/hooks/cogneris for processed documents, write the receiver that verifies the signature, and send a test delivery.

What the assistant does →

More, including your own classifiers and usage reports, in the prompt cookbook.

Why the code comes out right​

  • It reads the real contract, not a guess. The server gives the assistant the Cogneris documentation and the API reference as tools — every route with its schemas, errors and ready-to-run Node, Python, C# and curl.
  • It acts on your tenant, safely. API keys are created in the sandbox unless you explicitly confirm production. Keys created this way never carry management scopes.
  • Secrets stay out of your code. A new API key or webhook signing secret is returned once, and the assistant is told to write it to your git-ignored .env — never into chat, code or commits.
  • Same platform, same audit trail. Every call goes to the same Cogneris services as the REST API and is recorded against your user.

Prefer to write the code yourself?​

The REST API is the same one the assistant calls.

Bank CAN — API: Cogneris document APIs: classify, extract and face match in one call Watch on YouTube